search

hammerlab / biokepi

Bioinformatics Ketrew Pipelines
Apache License 2.0
27 stars 4 forks source link

Make biokepi use master branch igvxml #424

Closed ihodes closed 7 years ago

smondet commented 7 years ago

I'm not sure it's a good idea to use for the "defaults."

You can force the master branch to be used in your biokepi machine.

ihodes commented 7 years ago

Should I just use a commit in this case (to use a newer version of igvxml?)

hammer commented 7 years ago

@ihodes why not just cut a release of igvxml?

ihodes commented 7 years ago

Because updating on opam is kind of a pain for no benefit right now, and I anticipate continuing to change igvxml (lots of features need to be added) quite a bit in the near future for other uses of it.

hammer commented 7 years ago

Now that we have real patients in the pipeline I'd prefer we rely on released versions of software that can be recorded to improve reproducibility.

I also feel like a lot of pain for new users is caused by pinning.

ihodes commented 7 years ago

Gotcha—this is totally separate from the user needing to pin anything or know about what's happening. Biokepi pins to a commit here, and the commit hash would be noted so it's entirely reproducible.

smondet commented 7 years ago

@ihodes

I anticipate continuing to change igvxml (lots of features need to be added)

for that the best is using your Biokepi.Machine.t to force the one you want.

I agree with @hammer the defaults should be proper releases (with opam-publish the release process is pretty fast also).

ihodes commented 7 years ago

for that the best is using your Biokepi.Machine.t to force the one you want.

This is something I want all users to have, since it's effectively broken.

opam-publish isn't fast: it requires manual review.

Is there any reason to not be okay using a git tag or commit? As far as I can tell, it 100% addresses any concerns either of you have?

smondet commented 7 years ago

This is something I want all users to have…

That's why it's called "Release" ;)

opam-publish isn't fast: it requires manual review.

Well fully-automated could only be wrong…

Is there any reason to not be okay using a git tag or commit? As far as I can tell, it 100% addresses any concerns either of you have?

A github-only release (glorified tag) works for me. A bit sad to not push to the mothership opam-repo.

ihodes commented 7 years ago

Cool, I'll push a tag out for this commit.

hammer commented 7 years ago

Please also try to release regularly to OPAM as well.

ihodes commented 7 years ago

@smondet back to you!